ar
X
iv
:a
str
o-
ph
/0
60
33
82
v1
1
5
M
ar
2
00
6
Accepted to ApJ
Preprint typeset using LATEX style emulateapj v. 6/22/04
DISCOVERY OF THE COOLEST EXTREME SUBDWARF
Adam J. Burgasser1
Massachusetts Institute of Technology, Kavli Institute for Astrophysics and Space Research, Building 37, Room 664B, 77 Massachusetts
Avenue, Cambridge, MA 02139; ajb@mit.edu
and
J. Davy Kirkpatrick
Infrared Processing and Analysis Center, M/S 100-22, California Institute of Technology, Pasadena, CA 91125; davy@ipac.caltech.edu
Accepted to ApJ
ABSTRACT
We report the discovery of LEHPM 2-59 as the coolest extreme M subdwarf (esdM) found to date.
Optical and near infrared spectroscopy demonstrate that this source is of later spectral type than the
esdM7 APMPM 0559-2903, with the presence of strong alkali lines (including Rb I), VO absorption at
7400 A˚ and H2O absorption at 1.4 µm. Current optical classification schemes yield a spectral type of
esdM8, making LEHPM 2-59 one of only two ultracool esdMs known. The substantial space velocity of
this object (Vgalactic ≈ −180 km s−1) identifies it as a halo star. Spectral model fits to the optical and
near infrared spectral data for this and four other late-type esdMs indicate that LEHPM 2-59 is the
coolest esdM currently known, with Teff = 2800-3000K and -1.5 . [M/H] . -2.0. Comparison of Teff
determinations for M dwarfs and esdMs based on spectral model fits from this study and the literature
demonstrate a divergence in Teff scales beyond spectral types ∼M5/esdM5, as large as 600-800 K by
types M8/esdM8. While this divergence is likely an artifact of the underlying classification scheme,
it may lead to systematic errors in the derived properties of intermediate metallicity subdwarfs. We
comment on the future of ultracool subdwarf classification, and suggest several ideas for addressing
shortcomings in current (largely extrapolated) schemes.
Subject headings: stars: chemically peculiar — stars: individual (LEHPM 2-59, APMPM 0559-2903)
— stars: low mass, brown dwarfs — subdwarfs
1. INTRODUCTION
Subdwarfs are metal deficient stars lying below the
stellar main sequence in optical color-magnitude dia-
grams (Kuiper 1939). Low mass subdwarfs typically ex-
hibit halo kinematics (〈V 〉 = −202 km s−1; Gizis 1997),
and are presumably relics of the early Galaxy, with ages
&10 Gyr. With their extremely long lifetimes (far in
excess of the age of the Universe), low mass subdwarfs
are important tracers of Galactic structure and chemical
enrichment history, and are representatives of the first
generations of star formation.
The optical and near infrared spectra of the
coolest known M-type subdwarfs, like their solar-
metallicity dwarf counterparts, are dominated by molec-
ular absorption, including bands of CO, TiO, AlH,
CaH, CrH, FeH, MgH and H2O (Mould & Hyland
1976; Bessell 1982; Liebert & Probst 1987; Gizis
1997; Leggett, Allard, & Hauschildt 1998; Leggett et al.
2000). Collision induced H2 absorption (Linsky 1969;
Saumon et al. 1994; Borysow, Jørgensen, & Zheng 1997)
is also a strong absorber around 2 µm (Mould & Hyland
1976; Leggett et al. 2000). Variations in elemental abun-
dances (i.e., metallicity) can modulate these molecular
signatures appreciably, affecting both the total molec-
ular opacity (and hence overall luminosity at a given
mass) and relative band strengths through differential
chemical abundance patterns and modified atmospheric
1 Visiting Astronomer at the Infrared Telescope Facility, which is
operated by the University of Hawaii under Cooperative Agreement
NCC 5-538 with the National Aeronautics and Space Administra-
tion, Office of Space Science, Planetary Astronomy Program.
chemistry. At optical wavelengths, metallicity effects in
subdwarf spectra are seen succinctly in the relative band-
strengths of metal oxides and metal hydrides; the former
are weaker and the latter stronger in lower metallicity
dwarfs (Bidelman & Smethells 1976; Mould & Hyland
1976; Cottrell 1978; Bessell 1982). Current optical clas-
sification schemes for M subdwarfs are generally tied to
the relative strengths of these bands. For example, the
most widely used scheme, that defined by Gizis (1997,
hereafter G97), divides metal-poor M stars into subdwarf
(sdM) and extreme subdwarf (esdM) classes based on the
relative strengths of CaH and TiO bands in the 6300-7200
A˚ spectral region. G97 and Gizis & Reid (1997) have
determined mean metallicities of [Fe/H] = −1.2±0.3 and
[Fe/H] = −2.0±0.5 for these two classes of metal-poor
dwarfs.
Halo stars exhibit large space velocities relative to
the Sun; hence, they are efficiently detected in proper
motion surveys. Indeed, most low mass subdwarfs
now known were originally identified in blue photo-
graphic plate proper motion surveys, particularly those
by Luyten (e.g., LHS and NLTT catalogs; Luyten
1979a,b; see also Bakos, Sahu & Ne´meth 2002 and Salim
& Gould 2003). With the digitization of the red op-
tical R- and I-band photographic plate UK Schmidt
SERC and AAO (Harley & Dawe 1981; Cannon 1984;
Morgan et al. 1992), ESO (West & Schuster 1982; West
1984) and Palomar (POSS-I, Abell et al. 1959; POSS-
II, Reid et al. 1991) sky surveys, new proper mo-
tion surveys have begun to identify even cooler ob-
jects, which emit more of their light at wavelengths
redward of the visual band. These surveys – includ-

2 Burgasser & Kirkpatrick
ing the Automated Plate Measuring machine Proper
Motion survey (Scholz et al. 2000, hereafter APMPM),
the Calan-ESO proper motion survey (Ruiz et al. 2001),
the SUPERBLINK survey (Le´pine, Shara, & Rich 2002;
Le´pine & Shara 2005) and the SuperCOSMOS Sky Sur-
vey (Hambly et al. 2001a,b,c, hereafter SSS) – and as-
sociated follow-up programs2 have pushed subdwarf dis-
coveries to the end of the M spectral class (Scholz et al.
2004a). Metal-poor analogs to even cooler L dwarfs
(Kirkpatrick et al. 1999) have been identified in the
SUPERBLINK survey (Le´pine, Rich, & Shara 2003a,
hereafter LRS03) and serendipitously (Burgasser et al.
2003a; Burgasser 2004) in the Two Micron All Sky
Survey (Skrutskie et al. 2006, hereafter 2MASS). These
low temperature ultracool subdwarfs, encompassing
objects with spectral types sdM7/esdM7 and later
(Burgasser, Kirkpatrick, & Le´pine 2005, see also Kirk-
patrick, Henry & Irwin 1997) provide new challenges for
atmospheric modeling and extend our sampling of the
halo population down to and below the hydrogen burn-
ing minimum mass (Burgasser et al. 2003a).
While several ultracool subdwarfs are now known to
exist, only one ultracool extreme subdwarf has been
found, the esdM7 APMPM 0559-2903 (Schweitzer et al.
1999). To seek out even cooler subdwarfs, we have ini-
tiated a program to obtain spectra of red proper mo-
tion stars selected from the Liverpool Edinburgh High
Proper Motion survey (Pokorny, Jones & Hambly 2003;
Pokorny et al. 2004, hereafter LEHPM). One of these
sources, LEHPM 2-59, appears to be a later-type esdM
than APMPM 0559-2903 based on both near infrared
and optical data, and we identify it as the latest-type
and coolest esdM found to date. In §2 we summarize the
selection of this source from the LEHPM survey. In §3
we describe near infrared and optical spectroscopic ob-
servations of this and four other late-type esdMs, and
describe their overall characteristics. In §4 we analyze
line strengths, radial velocities and spectral types for the
optical data, and estimate the distance and kinematics of
LEHPM 2-59. In §5 we use spectral model fits to the ob-
served optical and near infrared data to derive Teff s and
metallicities for the latest-type esdMs. We discuss these
results in §6, focusing on the temperature scale of esdMs
and future revisions to existing classification schemes for
ultracool subdwarfs. Results are summarized in §7.
2. SELECTION OF LEHPM 2-59
The LEHPM catalog is a subset of 11289 proper mo-
tion stars from the SSS detected in R-band ESO and
UK Schmidt plates covering 7000 deg2 of the Southern
sky (δ . −20◦). This area excludes regions close to the
Galactic plane and those fields with epoch differences less
than 3 yr (the mean epoch difference is 8.5 yr). Catalog
sources were selected to have 0.′′18 yr−1 ≤ µ ≤ 20.′′0 yr−1,
µ/σµ > 3 and 9 ≤ R ≤ 19.5, and were cross-matched
with photographic BJ and IN plates (via the SSS) and
the 2MASS point source catalog. Further details on the
construction and completeness of the LEHPM catalog
2 See Reyle´ et al. (2002); Le´pine, Rich, & Shara (2003a);
Le´pine, Shara, & Rich (2004, 2003); Rojo & Ruiz (2003);
Pokorny, Jones & Hambly (2003); Pokorny et al. (2004);
Hambly et al. (2004); Reyle´ & Robin (2004); Scholz et al.
(2004a,b); Deacon, Hambly & Cooke (2005); Lodieu et al. (2005);
and Subasavage et al. (2005a,b).
are given in Pokorny et al. (2004).
We selected ultracool dwarf and subdwarf candidates
from the LEHPM catalog based on red optical/near-
infrared color and J-band reduced proper motion (RPM;
Hertzsprung 1905; Luyten 1922), HJ ≡ J+5 log10 µ+5 =
MJ +5 log10 Vtan − 3.73. RPM provides a measure of an
object’s absolute brightness and tangential space velocity
(Vtan) independent of its (generally unknown) distance,
and therefore enables the identification of intrinsically
faint and/or halo stars. Optical/near infrared RPM dia-
grams have become increasingly popular in searches for
ultracool nearby and halo stars (Salim & Gould 2002),
as these low temperature sources emit more of their flux
outside of the optical bands. Here, we focus on the J-
band RPM since the spectral energy distributions of late-
type dwarfs peak at these wavelengths. Figure 1 displays
the HJ versus (RESO−J) diagram for the LEHPM cata-
log. The main cluster of sources running diagonally down
the middle of the diagram is composed primarily of main
sequence dwarfs. The smaller cluster of sources offset
below and to the left is composed of metal-poor (shifting
their colors toward the blue), high velocity (increasing
HJ) halo subdwarfs. The smallest grouping in the lower
left corner of the diagram is primarily composed of low
luminosity white dwarfs.
To select for the latest-type sdMs and esdMs, we ap-
plied the following selection criteria to the LEHPM cat-
alog:
• Detection in both RESO and J bands, and
• (RESO − J) ≥ 1.5 and HJ ≥ 19.25 or
• (RESO−J) ≥ 3.5 and HJ ≥ 24.5−1.5(RESO−J).
The color/RPM criteria are illustrated in Figure 1. Im-
posing (RESO − J) ≥ 1.5 eliminates contamination by
the (relatively few) cool white dwarfs; the remaining cri-
teria cordon off the late-type extensions of the dwarf and
subdwarf tracks. A total of 50 sources were selected in
this manner, including LEHPM 2-59. Of these, 14 have
been previously observed by other programs, including 7
M8-M9 dwarfs (Gizis 2002; Lodieu et al. 2005), the sdM7
SSSPM J1930-4311 (Scholz et al. 2004a, a.k.a. LEHPM
2-31) and the esdM6 SSSPM J0500-5406 (Lodieu et al.
2005, a.k.a. LEHPM 1-3861). Results for our complete
LEHPM sub-sample will be presented in a forthcoming
publication.
3. OBSERVATIONS
3.1. Near Infrared Spectroscopy
LEHPM 2-59 and the three late-type esdMs LP
589-7 (Gizis & Reid 1999, esdM5), LSR 0822+1700
(Le´pine, Shara, & Rich 2004, esdM6.5) and APMPM
0559-2903 (Schweitzer et al. 1999, esdM7) were each ob-
served with the SpeX spectrograph (Rayner et al. 2003),
mounted on the 3m Infrared Telescope Facility, during
three runs on 2004 March 10, 2004 September 5–9 and
2005 December 31 (UT). A log of observations is provided
in Table 1. Conditions during these runs ranged from
poor (cloudy and high humidity) during 2004 March, to
clear with light cirrus during the 2004 September and
2005 December runs. Seeing was typically 0.′′5–0.′′9 on all
nights. Spectral data were obtained using the prism dis-
persed mode, which provides low resolution 0.7–2.5 µm

LEHPM 2-59 3
spectra in a single order. Using a 0.′′5 slit, we obtained
data with spectral resolution λ/∆λ ≈ 150 and dispersion
across the chip of 20–30 A˚ pixel−1. The slit was oriented
to the parallactic angle in all observations to reduce dif-
ferential color refraction, and the telescope was guided
by spillover light from the targets in the imaging chan-
nel. Multiple exposures of 180 s each were obtained in an
ABBA dither pattern along the slit. Nearby A0 V stars
were observed immediately after the target observations
and at a similar airmass (∆z < 0.1) for flux calibration
and telluric corrections. Internal flat field and Ar arc
lamps were also observed with each target for pixel re-
sponse and wavelength calibration.
All spectral data were reduced using the SpeXtool
package, version 3.3 (Cushing, Vacca, & Rayner 2004)
using standard settings. First, spectral images were cor-
rected for linearity, pair-wise subtracted, and divided
by the corresponding median-combined flat field image.
Spectra were then optimally extracted using default set-
tings for aperture and background source regions, and
wavelength calibration was determined from arc lamp
and sky emission lines. Individual spectral observations
for each science target and A0 star were normalized and
combined using a robust weighted mean with 8σ out-
lier rejection. Data for the A0 stars were also corrected
for broad-band shape variations before combining. Tel-
luric and instrumental response corrections for the sci-
ence data were determined using the method outlined in
Vacca et al. (2003), with line shape kernels derived from
arc lines. Adjustments were made to the telluric spectra
to compensate for differing H I line strengths in the ob-
served A0 V spectrum and pseudo-velocity shifts. Final
calibration was made by multiplying the observed target
spectrum by the telluric correction spectrum, which in-
cludes instrumental response correction through the ratio
of the observed A0 V spectrum to a scaled, shifted and
deconvolved Kurucz3 model spectrum of Vega.
The reduced near infrared spectra of the four esdMs are
shown in Figure 2. These spectra exhibit the characteris-
tic signatures of late-type esdMs, including blue near in-
frared spectral slopes due to strong collision-induced H2
absorption; shallow H2O absorption at 1.4 and 1.9 µm;
FeH and CrH absorption at 0.86, 0.99 and 1.2 µm; weak
bands of TiO at 0.84 µm; and K I and Na I lines at 0.77,
0.82 and 1.17 µm (Mould & Hyland 1976; Leggett et al.
2000). Notably absent are the deep H2O and 2.3 µm
CO bands that characterize solar-metallicity late-type
M dwarfs (Baldwin, Frogel & Persson 1973; Jones et al.
1994, cf. Fig. 1 of Burgasser et al. 2004). The weakness of
these bands makes near infrared classification of esdMs
difficult (Burgasser et al., in prep.; see §6.2); however,
the overall similarity of the spectrum of LEHPM 2-59 to
those of LSR 0822+1700 and APMPM 0559-2903, cou-
pled with its stronger H2O absorption and bluer spectral
slope, strongly suggests a late esdM spectral type.
3.2. Optical Spectroscopy
We obtained optical spectroscopy for LEHPM 2-59 and
the three late-type esdMs LP 589-7, SSSPM 0500-5406
and APMPM 0559-2903 on 2005 December 4 (UT) us-
ing the Low Dispersion Survey Spectrograph (LDSS-3)
mounted on the Magellan 6.5m Clay Telescope. A log of
3 http://kurucz.harvard.edu/stars.html .
observations is given in Table 2. LDSS-3 is an imaging
spectrograph, upgraded by M. Gladders from the original
LDSS-2 (Allington-Smith et al. 1994) for improved red
sensitivity. The instrument is composed of an STA0500A
4K×4K CCD camera that re-images an 8.′3 diameter field
of view at a pixel scale of 0.′′189. A set of slit masks and
grisms allow spectral observations at various resolutions
across the optical and red optical band. For our observa-
tions, we employed the VPH-red grism (660 lines/mm)
with a 0.′′75 (4 pixels) wide longslit to obtain 6050–10500
A˚ spectra across the entire chip with an average reso-
lution λ/∆λ ≈ 1800. Dispersion along the chip was 1.2
A˚/pixel. The OG590 longpass filter was used to eliminate
second order light shortward of 6000 A˚. Two long expo-
sures were obtained for each target, followed immediately
by a series of NeArHe arc lamp and flat-field quartz lamp
exposures. We also observed a nearby G2-G3 V star after
each esdM target for telluric absorption correction.
Data were reduced in the IRAF4 environment. We
focus our analysis on the blue side of LDSS-3 (in two-
amplifier readout mode), covering the 6050–8400 A˚ re-
gion. Raw science images were trimmed and subtracted
by a median combined set of bias frames taken during
the afternoon. The resulting images were divided by the
corresponding normalized, median-combined and bias-
subtracted set of flat field frames. Spectra were then
extracted using the APALL task with background sub-
traction but without variance weighting (i.e., not “opti-
mal extraction”). The dispersion solution for each target
was determined using the tasks REFSPEC, IDENTIFY
and DISPCOR, and arc lamp spectra extracted using
the same dispersion trace; solutions were typically accu-
rate to 0.05 pixels, or 0.07 A˚. Flux calibration was de-
termined using the tasks STANDARD and SENSFUNC
with observations of the spectral standard Hiltner 600
(Hamuy et al. 1994, a.k.a. HD 289002) obtained on 2005
December 3 (UT) with the same slit and grism combi-
nation as the science data. Corrections to telluric O2
(6850–6900 A˚ B-band, 7575–7700 A˚ A-band) and H2O
(7150–7300, 8150–8350 A˚) absorption for each esdM/G
star pair were determined by linearly interpolating over
these features in the G star spectrum, dividing by the
uncorrected G star spectrum, and multiplying the result
with the esdM spectrum. These corrections were gener-
ally adequate, but noticeable residuals are seen around
the O2 A-band in the spectra of SSSPM 0500-5406 and
APMPM 0559-2903, for which G star telluric calibra-
tors were observed at a larger differential airmass. These
residuals principally affect the blue wing of the K I dou-
blet at 7665/7699 A˚.
Reduced spectra for the four esdMs observed are shown
in Figure 3. Each spectrum shows characteristic spectral
traits of late-type esdMs, including weak TiO absorp-
tion bands at 6400, 6700 and 7150 A˚ (and possibly weak
TiO absorption at 7800 A˚); strong CaH bands at 6400
and 6900 A˚; and prominent line absorption from K I
(7665/7699 A˚ doublet) and Na I (8183/8195 A˚ doublet).
The K I doublet lines show substantial broadening in the
4 IRAF is distributed by the National Optical Astronomy Ob-
servatories, which are operated by the Association of Universities
for Research in Astronomy, Inc., under cooperative agreement with
the National Science Foundation.

4 Burgasser & Kirkpatrick
spectrum of LEHPM 2-59, reminiscent of the K I broad-
ening observed in solar metallicity L dwarfs. The 6708 A˚
Li I line is not detected in any of the spectra; this species
has almost certainly been depleted by fusion reactions in
the cores of all of these objects. There is a dip in the
spectra of SSSPM 0500-5406, APMPM 0559-2903 and
LEHPM 2-59 at 7400 A˚ which we attribute to the 1-0
B4Π-X4Σ− VO band, a feature commonly observed in
solar-metallicity M dwarf and early L dwarf spectra but
identified here for the first time in esdM spectra. This de-
tection is robust despite poor telluric O2 correction in the
spectra of SSSPM 0500-5406 and APMPM 0559-2903, as
this calibration error affects only the 7600–7700 A˚ win-
dow, while a dip in the spectrum is seen clearly at 7400–
7500 A˚ (see also § 5). Using the Kurucz Atomic Line
Database5 (Kurucz & Bell 1995), we have also identified
a forest of Ca I lines between 6100–6170 and 6440–6500
A˚, as well as more prominent lines at 6573 and 7326 A˚.
All of the atomic features identified in these spectra are
listed in Table 3; molecular line identifications can be
found in Kirkpatrick, Henry, & McCarthy (1991).
Rb I lines at 7800 and 7948 A˚ are also present in the
spectra of APMPM 0559-2903 and LEHPM 2-59, and the
7948 A˚ line is weakly present in the spectra of LP 589-7
and SSSPM 0500-5406. These lines are particularly in-
teresting, as they have been previously observed only in
the low resolution spectra of L dwarfs (Kirkpatrick et al.
1999) and, more recently, of the esdM6.5 LSR 0822-1700
(Le´pine, Shara, & Rich 2004) and the L subdwarf LSR
1610-0040 (Le´pine, Rich, & Shara 2003b, however, see
Cushing & Vacca 2006; Reiners & Basri 2006). On the
other hand, Rb I is not detected in low resolution spectra
of the latest-type M dwarfs or sdMs (Kirkpatrick et al.
1999; Le´pine, Shara, & Rich 2003; Scholz et al. 2004a).
This atomic species is largely ionized in the photospheres
of M dwarfs (Tion ∼ 2350 K; Lodders 1999, 2002), al-
though a significant fraction of Rb I gas is probably neu-
tral at higher temperatures in the higher pressure photo-
spheres of subdwarfs. The appearance of both the 7400
A˚ VO band and the 7800/7948 A˚ Rb I doublet lines could
be considered a defining trait of ultracool esdMs.
Finally, we note that Hα emission, a feature arising
from magnetic activity that is nearly always present
in the spectra of late-type M dwarfs (Gizis et al. 2000;
Hawley et al. 2002; West et al. 2004), is not detected in
any of these spectra (including LP 589-7; Gizis & Reid
1999), consistent with weak or absent chromospheres.
4. ANALYSIS
4.1. Atomic Line Strengths and Radial Velocities
Equivalent widths (EWs) for the alkali lines and the
two most prominent Ca I lines were measured using the
IRAF SPLOT routine. Uncertainties were determined as
the scatter of values from multiple measurements, while
upper limits for non-detected lines (Rb I in LP 589-7 and
SSSPM 0500-5406) were estimated from the mean EWs
of local noise features. These values are given in Table 4.
Note that EWs given for the K I and Na I doublets are
the combined absorption from both lines in each feature.
Nearly all of these lines are seen to strengthen from LP
5 Obtained through the online database search form
created by C. Heise and maintained by P. Smith; see
http://cfa-www.harvard.edu/amdata/ampdata/kurucz23/sekur.html .
589-7 to LEHPM 2-59, with the possible exception of the
7326 A˚ Ca I line, a transition that has the highest lower
energy state in Table 3 (4.6 eV; Kurucz & Peytremann
1988). The strengthening of atomic lines is observed in
the spectra of solar-metallicity M and L dwarfs as one
progresses to later spectral types, and is attributed to
lower photospheric temperatures and the corresponding
increase in column abundances of these neutral species.
The evolution of the line strengths in the esdMs observed
here also suggests a trend of cooler Teff with later esdM
type.
The six alkali lines and two strongest Ca I lines were
used to measure radial velocities (Vrad) for the observed
sources. Line centers were determined by gaussian fits
to the line cores, and velocity shifts were determined rel-
ative to the vacuum wavelengths listed in the Kurucz
database (Table 3). The mean and standard deviations
of these shifts are listed in Table 4, and include a sys-
tematic uncertainty of 3 km s−1 based on the mean accu-
racy of the dispersion solutions. Our values for LP 589-7,
SSSPM 0500-5406 and APMPM 0559-2903 are in agree-
ment with those of Gizis & Reid (1999); Lodieu et al.
(2005); and Schweitzer et al. (1999), respectively. With
Vrad = 79±9 km s−1, LEHPM 2-59 also appears to have
kinematics consistent with a halo subdwarf.
4.2. Spectral Classification
The optical spectra were classified using the scheme
of G97 as updated by LRS03. This scheme is based
on the relative strengths of the 7100 A˚ TiO band
and the 6400/6900 A˚ CaH bands, as sampled by the
four indices CaH1, CaH2, CaH3 and TiO5 defined in
Reid, Hawley, & Gizis (1995); as well as the redness
of the pseudo-continuum in the 6500–8200 A˚ spectral
region as sampled by the Color-M index of LRS03.
These indices were measured for each of the spectra af-
ter shifting them to their rest frame velocities. Fig-
ure 4 compares the combined CaH2+CaH3 indices to
TiO5 for these sources, in addition to measurements
for late-type dwarfs from Hawley, Gizis, & Reid (1996);
G97; Gizis & Reid (1997); Reid & Gizis (2005); LRS03;
Le´pine, Shara, & Rich (2004); and Scholz et al. (2004a).
This index-index diagram, based on work by G97, has
been used by several groups to segregate M dwarf spectra
by metallicity class. We update the divisions originally
set forth by G97 for the combined CaH2+CaH3 indices:
sdM : (CaH2 + CaH3) < 1.31(TiO5)3 −
2.37(TiO5)2 + 2.66(TiO5)− 0.20 (1)
esdM : (CaH2 + CaH3) < 3.54(TiO5)3 −
5.94(TiO5)2 + 5.18(TiO5)− 1.03. (2)
We stress that, as in G97, these divisions are somewhat
arbitrary as sources spanning Figure 4 represent a broad
continuum of metallicities. Figure 4 nevertheless demon-
strates that SSSPM 0500-5409, APMPM 0559-2903 and
LEHPM 2-59 are all bona-fide esdMs lying at the tail
end of the esdM distribution. LP 589-7 appears to be a
borderline sd/esd object; however, for the remainder of
this paper we will treat it as an esdM.
Numerical subtypes for the four observed sources were
determined using the relations
SpTesdM = 7.91(CaH2)
2 − 20.63(CaH2) + 10.71 (3)

LEHPM 2-59 5
SpTesdM = −13.47(CaH3) + 11.50 (4)
(i.e., Eqns. 2 and 8 from G97), and
SpTesdM = 22.0 ln(Color−M)− 4.1 (5)
(i.e., Eqn. 146 in LRS03), where SpTesdM (esdM0) = 0,
SpTesdM (esdM5) = 5, etc. Technically, these relations
are defined only up to subtype esdM6 (G97), but we fol-
low current practice (Le´pine, Shara, & Rich 2003, 2004;
Scholz et al. 2004a; Lodieu et al. 2005) in extrapolating
these relations to later numerical types (however, see
§6.2). Index values, index subtypes and averages of the
numerical subtypes (rounded off to the nearest 0.5 sub-
type) are listed in Table 5.
The derived subtypes for LP 589-7 and APMPM
0559-2903 are consistent with previous determinations.
We derive an esdM6.5 classification for SSSPM 0500-
5406, 0.5 subtypes later than the classification given by
Lodieu et al. (2005) who measured CaH2 and CaH3 in-
dices 0.05 larger than our values. However, this is gen-
erally consistent with the typical uncertainties in index
measurements and the corresponding uncertainties in de-
rived spectral types (±0.5 subtype). For LEHPM 2-59 we
derive a classification of esdM8, a full subtype later than
APMPM 0559-2903, until now the latest esdM known.
This result is consistent with the observed spectral trends
noted above, particularly the deeper near infrared H2O
bands and stronger optical atomic lines in the spectrum
of LEHPM 2-59. We conclude that LEHPM 2-59 is the
latest-type esdM found to date.
4.3. Distance and Space Velocity Estimates
There are currently few late-type esdMs with measured
parallaxes; only two sources esdM5 and later (the esdM5
LHS 3061 and the esdM6 LHS 1742a; Monet et al. 1992)
have known distances. Hence, only a rough distance es-
timate (dest) for LEHPM 2-59 is possible. For this, we
employed the spectral type/absolute magnitude relations
of LRS03 (Eqns. 24 and 25), which can be rewritten as
MR = 10.44 + 0.49(SpTesdM ) (6)
MKs = 8.06 + 0.37(SpTesdM ), (7)
where R is from USNO-B1.0 (Monet et al. 2003) and Ks
is from 2MASS. Reid & Gizis (2005) derived a similar
relation for MKs versus spectral type for esdMs,
MKs = 8.73 + 0.31(SpTesdM ). (8)
Assuming SpTesdM = 8, R = 18.82 and Ks = 14.76 (Ta-
ble 7), these three relations yield distances of 79, 55 and
51 pc, respectively. Gizis & Reid (1999) have also de-
rived absolute magnitude/color relations for esdMs; in
particular,
MI = 5.96 + 4.29(R− I), (9)
where R and I are photographic magnitudes. Again,
adopting photometry from USNO-B1.0 (R = 19.92, I
= 17.24) yields dest = 79 pc, in agreement with the
MR/spectral type relation in Eqn. 6. We therefore adopt
an average dest = 66±15 pc for this source, but stress
that such estimates are highly uncertain given the cur-
rent paucity of parallaxes for late-type subdwarfs.
6 Note that the LRS03 Color-M/spectral type relation uses the
natural logarithm of the Color-M index, and not the base-10 loga-
rithm as suggested by their Eqn. 14.
Using this estimated distance, and the measured
proper motion and radial velocity of LEHPM 2-59, we
computed UVW space velocity components relative to
the Local Standard of Rest (LSR). We derive [U, V,W ]
≈ [135,-180,-80] km s−1, assuming an LSR solar motion
of [U⊙, V⊙,W⊙] = [10,5,7] km s−1 (Dehnen & Binney
1998). These velocities lie well outside of the velocity
dispersion sphere of local disk M dwarfs ([σU ,σV ,σW ] ≈
[40,28,19] km s−1 centered at [-13,-23,-7] km s−1; Haw-
ley, Gizis & Reid 1996); and the large negative V veloc-
ity, ranging over -210 to -145 km s−1 for our distance
estimates, is consistent with motion independent of the
Galactic disk. Assuming an LSR rest frame velocity of
-220 km s−1 (Kerr & Lynden-Bell 1986), the total space
velocity of LEHPM 2-59 in the Galactic potential is 135–
190 km s−1. This is well below the Galactic escape veloc-
ity in the vicinity of the Sun (Carney, Latham, & Laird
1988; Leonard & Tremaine 1990). Interestingly, the or-
bit of this object is highly elliptical, and passes well inside
the Galactic bulge on its closest approach to the Galactic
center. This is not atypical for halo stars currently in the
vicinity of the Sun (S. Le´pine, 2006, private communica-
tion).
5. SPECTRAL MODEL FITS
To further gauge the physical properties of these late-
type esdMs, we compared our optical and near infrared
spectra to subsolar metallicity theoretical spectral mod-
els from Hauschildt, Allard & Baron (1999, NextGen)
and Allard et al. (2001, AMES Cond). These models are
based on the Phoenix code (Hauschildt, Baron & Allard
1997, and references therein), employ self-consistent tem-
perature/pressure profiles, and assume local thermody-
namic equilibrium. Both sets of models use the opac-
ity sampling method; a full account of the chemical
species and opacities used by these models is provided
in Allard et al. (2001) and references therein. Two ma-
jor changes in the Cond models are the removal of con-
densed species from the upper atmosphere and the use of
updated TiO and H2O opacities. The NextGen models
and their antecedents (Allard 1990; Allard & Hauschildt
1995; Allard et al. 1997) have been used extensively for
fitting M subdwarf spectra (Gizis 1997; Schweitzer et al.
1999; Dawson & De Robertis 2000; Leggett et al. 2000;
Le´pine, Shara, & Rich 2004).
We sampled grids of NextGen and Cond models span-
ning temperatures of 2600 ≤ Teff≤ 3600 K in steps of
100 K, metallicities of -3.0 ≤ [M/H] ≤ 0.0 dex in steps
of 0.5 dex, and surface gravities of log g = 5.0 and 5.5
cm s−2. Comparisons were made separately to the op-
tical and near infrared data. For the optical fits, both
empirical and model spectra were normalized at 8100 A˚
as in Figure 3, and the observed data were shifted to
their rest frame velocities. For the near infrared fits,
data and models were scaled to 1.2 µm. Model spectra
were also deconvolved to the resolution of the observed
data (λ/∆λ = 1800 and 150 for the optical and near in-
frared, respectively) using a Gaussian kernel. For each
spectrum/model pairing, the root mean square (RMS)
deviation was computed over the spectral ranges 6050–
7500 and 7650–8250 A˚ in the optical (to avoid regions of
poor telluric O2 correction and the 7400 A˚ VO band; see
below); and 0.75–1.31, 1.45–1.75 and 2.0–2.4 µm in the
near infrared (to avoid telluric H2O regions). The nor-

6 Burgasser & Kirkpatrick
malization of the model spectra in each fit was allowed
to vary by a factor ranging from 0.75 to 1.25 (in steps
of 0.05) to allow for continuum offsets, and the normal-
ization with the minimum RMS was retained. For each
model set and surface gravity, the Teff and [M/H] com-
bination with the overall minimum RMS was deemed to
be the best fit for a particular spectrum. These best fit
values are listed in Table 6.
Figure 5 and 6 compares the best fit NextGen and
Cond, log g = 5.0 models for the spectrum of LEHPM
2-59 in the optical and near infrared, respectively. Over-
all, these fits are reasonably good, with the broad spec-
tral shape and several individual features matching quite
well. For the optical data, the log g = 5.0 NextGen mod-
els consistently provided the best fits to the data. This
appears to be largely due to better fits to the 6700 A˚ CaH
band which is only slightly overestimated in the NextGen
models but greatly underestimated in the Cond mod-
els. The 7150 A˚ TiO band is too strong in both model
sets; higher Teff and/or lower [M/H] models weaken this
band but show very poor agreement with the rest of the
optical spectrum. TiO features have long been identi-
fied as a problem, particularly in the NextGen models
(Allard, Hauschildt & Schwenke 2000). Atomic lines are
also generally too strong in the best fit models. Indeed,
there are many metal lines apparent in the Cond mod-
els for which no analogs are seen in the empirical data.
The Na I, Rb I and prominent Ca I lines are also too
deep, although the K I doublet is reproduced quite well,
both in depth and breadth, for both sets of models. The
apparent mismatch over 7300-7500 A˚ is due to the 7400
A˚ VO band, which is not included in the opacity set of
either the NextGen or Cond models (Allard et al. 2001).
We found slight systematic effects in the derived param-
eters between different model sets and surface gravities.
The best fit Cond models yielded slightly lower Teff s
(by . 100 K) and higher [M/H] values (by . 0.5 dex) as
compared to the best fit NextGen models; higher gravity
model fits typically gave higher Teff s (by . 100 K) and
[M/H] values (by . 0.5 dex).
In the near infrared, spectral model fits again provided
reasonably good matches to the broad spectral energy
distribution, with slight discrepancies in the 0.99 µm FeH
band, the 1.15–1.30 µm continuum (also dominated by
FeH absorption in late-type subdwarfs; Burgasser et al.
2004, Cushing et al. 2006; Burgasser et al. in prep.) and
the 1.4 µm H2O band. In this wavelength regime, Cond
models typically provided the best fits, largely due to
better agreement in the 0.8–1.1 µm region. Systematic
shifts in the derived parameters between the NextGen
and Cond model sets were less pronounced at these wave-
lengths, but higher surface gravity models consistently
gave larger Teff s (by 100–200 K) and slightly lower
[M/H] values (by . 0.5 dex). There was also some degen-
eracy in the best fit values, with hotter, solar metallic-
ity models providing fairly good matches to the data for
λ > 1.1µm. This is not surprising, given that the blue,
relatively featureless near infrared spectra of esdMs are
not unlike those of hotter M and K dwarf stars.
Derived parameters between the optical and near in-
frared fits for a given object exhibit clear systematic dif-
ferences. Teff s derived from the optical data are gener-
ally 100-200 K higher than those derived from the near
infrared data; metallicities are consistently 0.5–1.0 dex
higher. Examination of the models indicates that lower
metallicities are required to match the H- and K-band
suppression in the near infrared data, but can result in
optical CaH bands that are too deep. Similar effects
are observed with lower Teff s. Since we have no way
of independently validating fits in either spectral region,
we must treat these systematic differences as a source of
uncertainty in the model fits, emphasizing that absolute
Teff s and [M/H] values derived from spectral model fits
should in general be treated with some caution.
Turning to the derived parameters of the objects in
our sample, we find that, overall, later esdM types cor-
respond to cooler Teff s, consistent with expectations.
The one notable exception is the esdM6.5 LSR 0822-
1700, for which we derive a temperature 100 K cooler
than the esdM7 APMPM 0559-2903. This slight dif-
ference may not be significant; indeed, our Teff deter-
mination based on near infrared data is 200 K cooler
than that derived by Le´pine, Shara, & Rich (2004) from
optical data. However, metallicity effects may be at
play. LSR 0822-1700 appears to have a consistently lower
[M/H] than APMPM 0559-2903 based on both optical
(Le´pine, Shara, & Rich 2004) and near infrared analy-
sis. In any case, LEHPM 2-59 is the lowest temperature
object in the group, with Teff ≈ 2800–3000 K. Evolu-
tionary models by Baraffe et al. (1997) predict a mass
of ∼0.09 M⊙ for these parameters assuming an age of
10 Gyr, only ∼0.01 M⊙ above the hydrogen burning min-
imum mass for these metallicities.
6. DISCUSSION
6.1. The Temperature Scales of M Dwarfs and Extreme
Subdwarfs
The temperatures derived for late-type esdMs in this
and other studies (Schweitzer et al. 1999; Leggett et al.
2000; Le´pine, Shara, & Rich 2004) are relatively warm
compared to solar-metallicity dwarfs with comparable
subtypes, which typically have Teff ∼ 2200–2800 K
(Kirkpatrick et al. 1993; Leggett, Allard, & Hauschildt
1998; Leggett et al. 2000; Golimowski et al. 2004). Fig-
ure 7 compares Teff determinations for M dwarfs and es-
dMs based on spectral model fits to NextGen models and
their antecedents. These data are segregated by optical7
(Kirkpatrick et al. 1993; Gizis 1997; Schweitzer et al.
1999; Le´pine, Shara, & Rich 2004) and near infrared
(Kirkpatrick et al. 1993; Leggett et al. 1996, 2000, 2001;
Dawson & De Robertis 2000) analyses. There is a sig-
nificant downturn in the dwarf Teff scale around type
M5–M6, dropping from ∼3000 K to ∼2200 K by type
M8. The esdMs, on the other hand, show a more grad-
ual decrease in Teff with spectral type over the same
range. Linear fits to optical and near infrared Teff de-
terminations over the range esdM1-esdM8 yield parallel
relations:
Teff = 3750− 93×SpTesdM K (optical) (10)
Teff = 3620− 96×SpTesdM K (near infrared) (11)
with dispersions of 30 and 90 K, respectively. The tem-
peratures of the two latest esdMs, APMPM 0559-2903
7 We did not include Teff determinations for M dwarfs later
than M6 in the Kirkpatrick et al. (1993) and G97 studies due to
the poor quality fits to ultracool dwarf spectra existing models
provided at that time.

LEHPM 2-59 7
and LEHPM 2-59, are 600–800 K hotter than those of
equivalently typed solar-metallicity dwarfs.
Why would the effective temperatures of ultracool es-
dMs differ so greatly from those of ultracool solar metal-
licity dwarfs? It is well known that lower metallicity
implies lower overall atmospheric opacity, a consequence
of reduced molecular and H− abundances. Hence, for a
given mass, the observed photosphere lies at deeper, hot-
ter and higher pressure layers in more metal-poor stars.
However, this argument serves only to explain (possi-
ble) differences in Teff/mass relations between metal-
licity classes. Teff/spectral type relations also hinge
on how the spectral types themselves are derived. In
fact, the classification of M dwarfs and esdMs are de-
termined by separate relations in the G97 and LRS03
schemes, as necessitated by the divergence of spectral
properties between these metallicity classes. Hence, dif-
ferences in the Teff/spectral type relations between M
dwarfs and esdMs may simply be an artifact of the un-
derlying classification scheme. At first glance, this may
seem unimportant, for as long as the correct scale is used
for a given metallicity class, one should derive the cor-
rect Teff . However, metallicity is not a discrete quantity
like metallicity class, and systematic deviations can occur
for individual objects that are more or less metal-poor
than the mean of their assigned class. A more global
Teff/metallicity/spectral type relation is required.
6.2. Future Prospects for Ultracool Subdwarf
Classification
With the recent compilation of several red optical
proper motion surveys and the initiation of near infrared
programs (Deacon, Hambly & Cooke 2005, Looper et al.,
in prep.), it is quite likely that LEHPM 2-59’s status as
the coolest esdM known will not be long-lived. Indeed,
the identification of a few L subdwarfs (sdL) to date sug-
gests that esdL discoveries may not not far off (however,
see below). It has been shown by Burgasser et al. (2003a)
and Le´pine, Rich, & Shara (2003b) that these ultracool
metal-poor stars and brown dwarfs cannot be adequately
classified by existing schemes largely due to the disap-
pearance of the 7150 A˚ TiO band. The lack of flux from
these cool objects also argues against classification in the
6300-7200 A˚ region.
So where do we proceed? An obvious option is to de-
fine new subdwarf classification schemes at longer wave-
lengths, including the red optical (cf. late-M and L
dwarfs) and near infrared (cf. T dwarfs) spectral regions.
Absorption bands from the metal hydrides FeH and CrH
(and perhaps TiH; Burrows et al. 2005) are prominent in
ultracool dwarf and subdwarf spectra for λ > 8000 A˚, and
are likely to be present in even lower metallicity analogs.
Contrasting these features with longer wavelength TiO
(8200 and 8400 A˚) and VO bands (7900, 9500 and 10500
A˚), alkali lines (Na I, K I, Rb I and Cs I) or pseudo-
continuum slope should prove to be effective diagnos-
tics for segregating metallicity and temperature classes.
At longer wavelengths, subdwarf classification becomes
more difficult due to the weakness of absorption features
and the strong suppression of H- and K-band flux by H2
absorption (which also wipes out the 2.3 µm CO band).
Nevertheless, ratios comparing the 1.4 µm H2O band and
the near infrared spectral slope might provide some dis-
crimination (Burgasser et al. in prep.), while metal line
features (including Al I; see Cushing & Vacca 2006) and
FeH bands in the 1-1.35 µm region could be used for
higher resolution studies.
As later extreme subdwarfs are identified, what will
define the termination of the esdM class and the begin-
ning of the esdLs? Solar-metallicity L dwarfs are dis-
tinguished by waning metal oxide bands, strengthening
metal hydride and alkali lines, steep red optical slopes
and red near infrared colors. Yet esdMs already exhibit
weak metal oxides and strong metal hydrides, and sev-
eral neutral alkali species (including Rb I and Cs I); while
near infrared colors will never become red due to strong
absorption by H2 (Saumon et al. 1994). Spectral mod-
els cannot provide clear guidance on this transition, as
it remains unclear as to whether condensate formation,
a key aspect of the M/L dwarf transition, plays a sig-
nificant role in low-temperature metal-poor atmospheres
(Burgasser et al. 2003a). Clearly, cooler extreme subd-
warf discoveries must guide the definition and our under-
standing of this transition.
One further complication in this issue is whether halo
esdLs even exist in our Galaxy today. As metal-poor
halo objects about the substellar mass limit evolve over
∼10 Gyr, wide gaps in the luminosities and effective
temperatures of this population develop. Low mass
stars attain a steady-state Teff & 2500-3000 K, while
most brown dwarfs cool to Teff . 1000 K in this time
(Burrows et al. 2001, cf. Figure 3 in Burgasser et al.
2003a). Only a narrow range of low mass star/brown
dwarf transition objects are expected to encompass in-
termediate L-type Teff s between these limits, and could
therefore be quite rare. Future surveys for low luminosity
metal-poor objects will hopefully probe this L dwarf gap,
providing, if little data for esdLs, a unique constraint for
brown dwarf evolutionary theories.
Finally, one area in which subdwarf classification can
be immediately improved is the identification of spe-
cific standard stars for temperature and metallicity sub-
types. A framework of standard stars is a fundamental
tenet of the MK Process (Morgan, Keenan & Kellman
1943; Morgan & Keenan 1973; Keenan & McNeil 1976;
Corbally, Gray & Garrison 1994), the most widely
adopted method of stellar classification. Standard
stars provide the basis of current classification schemes
for solar metallicity M (Kirkpatrick, Henry, & McCarthy
1991), L (Kirkpatrick et al. 1999) and T dwarfs
(Burgasser et al. 2006). Allowing specific stars to de-
fine a classification scheme provides a level of consistency
that is not generally present in pure index schemes, while
retaining independence from constantly evolving theo-
retical interpretations. There are now several dozen M
subdwarfs of types sdM5/esdM5 and earlier from which
appropriate standards can be chosen, and we anticipate
that current red optical and near infrared proper motion
surveys will soon fill in the remainder of the M subdwarf
and extreme subdwarf sequences. Now is an opportune
time to consider the construction of a more robust clas-
sification scheme for late-type subdwarfs.
7. SUMMARY
We have identified LEHPM 2-59 as the coolest esdM
identified to date. Near infrared and optical spectroscopy
show features indicative of a late-type esdM, including

8 Burgasser & Kirkpatrick
strong alkali lines, VO and H2O absorption and a blue
near infrared spectral slope. We derive an optical spec-
tral type of esdM8 on the G97 and LRS03 schemes, and
spectral model fits to both optical and near infrared data
indicate Teff = 2800–3000 K and -1.5 . [M/H] . -2.0 for
this source. Its kinematics confirm it as a halo star. We
show that the temperatures of this and other late-type
esdMs are significantly higher than equivalently classi-
fied solar-metallicity M dwarfs, a divergence that could
lead to systematic errors in derived parameters for ob-
jects with intermediate metallicities. Finally, we have
touched on the future prospects of M subdwarf classifi-
cation, and methods by which it may be improved and
extended to later subtypes. With the current high dis-
covery rate of low-temperature, ultracool halo subdwarfs,
it is likely that these issues will require further scrutiny
in the near term.
We thank P. Hauschildt and F. Allard for making
their team’s spectral models electronically available, and
acknowledge useful discussions with P. Hauschildt, S.
Le´pine, S. Mohanty and J. Mulchaey during the prepa-
ration of the manuscript. We also thank our anonymous
referee and our scientific editor, J. Liebert, for their help-
ful comments. AJB acknowledges the assistance of tele-
scope operators B. Golisch, D. Griep and P. Sears at
IRTF, and H. Rivera and S. Vera at Magellan during the
observations presented in this study, as well as our instru-
ment scientists J. Rayner (SpeX) and J. Bravo (LDSS-3).
This publication makes use of data from the Two Micron
All Sky Survey, which is a joint project of the University
of Massachusetts and the Infrared Processing and Anal-
ysis Center, and funded by the National Aeronautics and
Space Administration and the National Science Founda-
tion. 2MASS data were obtained from the NASA/IPAC
Infrared Science Archive, which is operated by the Jet
Propulsion Laboratory, California Institute of Technol-
ogy, under contract with the National Aeronautics and
Space Administration. The authors wish to extend spe-
cial thanks to those of Hawaiian ancestry on whose sacred
mountain we are privileged to be guests.
Facilities: IRTF (SpeX); Magellan (LDSS-3)
REFERENCES
Abell, G. O. 1959, PASP, 67, 258
Allard, F. 1990, Ph.D. Thesis, Ruprecht-Karls University,
Heidelberg
Allard, F., & Hauschildt, P. H., 1995, ApJ, 445, 433
Allard, F., Hauschildt, P. H., Alexander, D. R., & Starrfield, S.
1997, ARA&A, 35, 137
Allard, F., Hauschildt, P. H., Alexander, D. R., Tamanai, A., &
Schweitzer, A. 2001, ApJ, 556, 357
Allard, F., Hauschildt, P. H., & Schwenke, D. 2000, ApJ, 540, 1005
Allington-Smith, J., et al. 1994, PASP, 106, 983
Bakos, G. A´., Sahu, K. C., & Ne´meth, P. 2002, ApJS, 141, 187
Baldwin, J. R., Frogel, J. A., & Persson, S. E. 1973, ApJ, 184, 427
Baraffe, I., Chabrier, G., Allard, F., & Hauschildt, P. H. 1997,
A&A, 327, 1054
Bessell, M. S. 1982, PASA, 4, 417
Bidelman, W. P., & Smethells, W. G. 1976, in IAU Symp. 72,
Abundance Effects in Classification ed. B. Hauck, P. C. Keenan
& W. W. Morgan (Boston: D. Reidel), p. 205
Borysow, A., Jørgensen, U. G., & Zheng, C. 1997, A&A, 324, 185
Burgasser, A. J. 2004 ApJ, 614, L73
Burgasser, A. J., Geballe, T. R., Leggett, S. K., Kirkpatrick, J. D.,
& Golimowski, D. A. 2006 ApJ, in press
Burgasser, A. J., Kirkpatrick, J. D., Burrows, A., Liebert, J., Reid,
I. N., Gizis, J. E., McGovern, M. R., Prato, L., & McLean, I. S.
2003a, ApJ, 592, 1186
Burgasser, A. J., Kirkpatrick, J. D., & Le´pine, S. 2005 in The
13th Cambridge Workshop on Cool Stars, Stellar Systems, and
the Sun (ESA-SP-560), ed. F. Favata, G. A. J. Hussain & B.
Battrick (Noordwijk: ESA), p. 237
Burgasser, A. J., Kirkpatrick, J. D., McElwain, M. W., Cutri, R.
M., Burgasser, A. J., & Skrutskie, M. F. 2003, AJ, 125, 850
Burgasser, A. J., McElwain, M. W., Kirkpatrick, J. D., Cruz, K.
L., Tinney, C. G., & Reid, I. N. 2004 AJ, 127, 2856
Burrows, A., Dulick, M., Bauschlicher, C. W., Jr., Bernath, P. F.,
Ram, R. S., Sharp, C. M., & Milsom, J. 2005, ApJ, 624, 988
Burrows, A., Hubbard, W. B., Lunine, J. I., & Liebert, J. 2001,
Rev. of Modern Physics, 73, 719
Cannon, R. D. 1984, in Astronomy with with Schmidt-Type
Telescopes, Proc. IAU Coll. 78, ed. M. Cappaccioli (Dordrecht:
Reidel), p. 25
Carney, B., Latham, D., & Laird, J. B. 1988, AJ, 96, 560
Corbally, C. J., Gray R. O., & Garrison, R. F. 1994, The MK
process at 50 years. A Powerful Tool for Astrophysical Insight
(San Francisco: Astronomical Society of the Pacific)
Cottrell, P. L. 1978, ApJ, 223, 544
Cushing, M. C., & Vacca, W. D. 2006, AJ, in press
Cushing, M. C., Vacca, W. D., & Rayner, J. T. 2004, PASP, 116,
362
Dawson, P. C., & De Robertis, M. M. 2000, AJ, 120, 1532
Deacon, N. R., Hambly, N. C., & Cooke, J. A. 2005, A&A, 435,
363
Dehnen, W., & Binney, J. J. 1998, MNRAS, 298, 387
Epchtein, N., et al. 1997, The Messenger, 87, 27
Gizis, J. E. 2002, ApJ, 575, 484
Gizis, J. E. 1997, AJ, 113, 806 (G97)
Gizis, J. E., Monet, D. G., Reid, I. N., Kirkpatrick, J. D., Liebert,
J., & Williams, R. 2000, AJ, 120, 1085
Gizis, J. E., & Reid, I. N. 1997, PASP, 109, 849
—. 1999, AJ, 117, 508
Golimowski, D. A., et al. 2004, AJ, 127, 3516
Hambly, N. C., Davenhall, A. C., Irwin, M. J., & MacGillivray, H.
T. 2001a, MNRAS 326, 1315
Hambly, N. C., Irwin, M. J., & MacGillivray, H. T. 2001b, MNRAS
326, 1295
Hambly, N. C., MacGillivray, H. T., Read, M. A., et al. 2001c,
MNRAS 326, 1279
Hambly, N. C., Henry, T. J., Subasavage, J. P., Brown, M. A., &
Jao, W.-C. 2004, AJ, 128, 437
Hamuy, M., Suntzeff, N. B., Heathcote, S. R., Walker, A. R.,
Gigoux, P., & Phillips, M. M. 1994, PASP, 106, 566
Hartley, M., & Dawe, J. A. 1981, PASA, 4, 251
Hauschildt, P. H., Allard, F., & Baron, E. 1999, ApJ, 512, 377
Hauschildt, P. H., Baron, E., & Allard, F. 1997, ApJ, 483, 390
Hawley, S. L., Gizis, J. E., & Reid, I. N. 1996, AJ, 112, 2799
Hawley, S. L. et al. 2002, AJ, 123, 3409
Hertzsprung, R. J. 1905, Zeit. Phot., 3, 429
Jones, H. R. A., Longmore, A. J., Jameson, R. F., & Mountain, C.
M. 1994, MNRAS, 267, 413
Keenan, P. C., & McNeil, R. C. 1976, An Atlas of Spectra of the
Cooler Stars: Types G, K, M, S and C (Columbus: Ohio State
Univ. Press)
Kerr, F. J., & Lynden-Bell, D. 1986, MNRAS, 221, 1023
Kirkpatrick, J. D., Henry, T. J., & Irwin, M. J. 1997, AJ, 113, 1421
Kirkpatrick, J. D., Henry, T. J., & McCarthy, D. W., Jr. 1991,
ApJS, 77, 417
Kirkpatrick, J. D., Kelly, D. M., Rieke, G. H., Liebert, J., Allard,
F., & Wehrse, R. 1993, ApJ, 402, 643
Kirkpatrick, J. D., et al. 1999, ApJ, 519, 802
Kuiper, G. P. 1939, ApJ, 89, 549
Kurucz, R. L. 1988, Trans. IAU, XXB, ed. M. McNally, (Dordrecht:
Kluwer), p. 168-172
Kurucz, R. L., & Bell, B. 1995, Atomic Line Data, Kurucz
CD-ROM No. 23. (Cambridge: Smithsonian Astrophysical
Observatory)
Kurucz, R. L., & Peytremann, E. 1975, SAO Special Report, 362
Leggett, S. K., Allard, F., Berriman, G., Dahn, C. C., & Hauschildt,
P. H. 1996, ApJS, 104, 117

LEHPM 2-59 9
Leggett, S. K., Allard, F., Dahn, C., Hauschildt, P. H., Kerr, T.
H., & Rayner, J. 2000, ApJ, 535, 965
Leggett, S. K., Allard, F., Geballe, T., Hauschildt, P. H., &
Schweitzer, A. 2001, ApJ, 548, 908
Leggett, S. K., Allard, F., & Hauschildt, P. H. 1998, ApJ, 509, 836
Leonard, P. J. T., & Tremaine, S. 1990, ApJ, 353, 486
Le´pine, S., Rich, R. M., & Shara, M. M. 2003a, AJ, 125, 1598
(LRS03)
Le´pine, S., Rich, R. M., & Shara, M. M. 2003b, ApJ, 591, L49
Le´pine, S., & Shara, M. M. 2005, AJ, 129, 1483
Le´pine, S., Shara, M. M., & Rich, R. M. 2002, AJ, 124, 1190
Le´pine, S., Shara, M. M., & Rich, R. M. 2003, ApJ, 585, L69
Le´pine, S., Shara, M. M., & Rich, R. M. 2004, ApJ, 602, L125
Liebert, J., & Probst, R. G. 1987, ARA&A, 25, 473
Linsky, J. L. 1969, ApJ, 156, 989
Lodders, K. 1999, ApJ, 519, 793
Lodders, K. 2002, ApJ, 577, 974
Lodieu, N., Scholz, R.-D., McCaughrean, M. J., Ibata, R., Irwin,
M., & Zinnecker, H. 2005, A&A, 440, 1061
Luyten, W. J. 1922, Lick Obs. Bull., 10, 135
Luyten, W. J. 1979a, LHS Catalogue: A Catalogue of Stars with
Proper Motions Exceeding 0.′′5 Annually (Minneapolis: Univ.
Minn. Press)
—. 1979b, New Luyten Catalogue of Stars with Proper Motions
Larger than Two Tenths of an Arcsecond (NLTT) (Minneapolis:
Univ. Minn. Press)
Monet, D. G., Dahn, C. C., Vrba, F. J., Harris, H. C., Pier, J. R.,
Luginbuhl, C. B., & Ables, H. D. 1992, AJ, 103, 638
Monet, D. G., et al. 2003, AJ, 125, 984 (USNO-B1.0 Catalog)
Morgan, W. W., & Keenan, P. C. 1973, ARA&A, 11, 29
Morgan, W. W., Keenan, P. C., & Kellman, E. 1943, An Atlas
of Stellar Spectra, with an Outline of Spectral Classification
(Chicago: Univ. Chicago Press)
Morgan, W. W., Sharpless, S., & Osterbrock, D. 1952, AJ, 57, 3
Morgan, D. H., Tritton, S. B., Savage, A., Hartley, M., & Cannon,
R. D. 1992, in Digitised Optical Sky Surveys, ed. H. T.
MacGillivray & E. B. Thomson (Dordrecht: Boston), p. 11
Mould, J. R., & Hyland, A. R. 1976, ApJ, 208, 399
Pokorny, R. S., Jones, H. R. A., & Hambly, N. C. 2003, A&A, 397,
575
Pokorny, R. S., Jones, H. R. A., Hambly, N. C., & Pinfield, D. J.
2004, A&A, 421, 763
Rayner, J. T., Toomey, D. W., Onaka, P. M., Denault, A. J.,
Stahlberger, W. E., Vacca, W. D., Cushing, M. C., & Wang,
S. 2003, PASP, 155, 362
Reid, I. N., & Hawley, S. L. 2000, New Light on Dark Stars
(Chichester: Praxis)
Reid, I. N., & Gizis, J. E. 2005, PASP, 117, 676
Reid, I. N., Hawley, S. L., & Gizis, J. E. 1995, AJ, 110, 1838
Reid, I. N., Kirkpatrick, J. D., Liebert, J., Gizis, J. E., Dahn, C.
C., & Monet, D. G. 2002, AJ, 124, 519
Reiners, A., & Basri, G. 2006, AJ, in press
Reyle´, C., & Robin, A. C. 2004, A&A, 421, 643
Reyle´, C., Robin, A. C., Scholz, R.-D., & Irwin, M. 2002, A&A,
390, 491
Rojo, P., & Ruiz, M. T. 2003, AJ, 126, 353
Ruiz, M. T., Wischnjewsky, M., Rojo, P. M., & Gonzalez, L. E.
2001, ApJS, 133, 119
Salim, S., & Gould, A. 2002, ApJ, 575, L83
Salim, S., & Gould, A. 2003, ApJ, 582, 1011
Saumon, D., Bergeron, P., Lunine, J. I., Hubbard, W. B., &
Burrows, A. 1994, ApJ, 424, 333
Scholz, R.-D., Irwin, M., Ibata, R., Jahreiβ, H., & Malkov, O. Yu.
2000, A&A, 353, 958
Scholz, R.-D., Lehmann, I., Matute, I., & Zinnecker, H. 2004, A&A,
425, 519
Scholz, R.-D., Lodieu, J., & McCaughrean, M. 2004, A&A, 428,
L25
Schweitzer, A., Shultz, R.-D., Stauffer, J., Irwin, M., &
McCaughren, M. J. 1999, A&A, 350, L62
Skrutskie, M. F., et al. 2006, AJ, 131, 1163
Subasavage, J. P., Henry, T. J., Hambly, N. C., Brown, M. A., &
Jao, W.-C. 2005a, AJ, 129, 413
Subasavage, J. P., Henry, T. J., Hambly, N. C., Brown, M. A., Jao,
W.-C., & Finch, C. T. 2005b, AJ, 130, 1658
Vacca, W. D., Cushing, M. C., & Rayner, J. T. 2003, PASP, 155,
389
Warner, B. 1968, MNRAS, 140, 53
West, A. A., et al. 2004, AJ, 128, 426
West, R. M. 1984, in Astronomy with with Schmidt-Type
Telescopes, Proc. IAU Coll. 78, ed. M. Cappaccioli (Dordrecht:
Reidel), p. 13
West, R. M., & Schuster, H.-E. 1982, A&AS, 49, 577
Wiese, W. L., Smith, M. W., & Glennon, B. M. 1966, Atomic
Transition Probabilities, Vol. 1. (Washington, D. C.: GPO)
York, D. G., et al. 2000, AJ, 120, 1579

10 Burgasser & Kirkpatrick
Fig. 1.— HJ versus (RESO −J) diagram for the LEHPM catalog (Pokorny et al. 2004). Note the three clusters of sources corresponding
to (from center to lower left) main sequence dwarfs, halo subdwarfs and white dwarfs. Our selection criteria for ultracool dwarf and
subdwarf candidates are indicated by dashed lines. Previously classified sources within this region are noted by open circles; LEHPM 2-59
is indicated by an open square. [THIS FIGURE IS INCLUDED IN JPEG FORMAT]
Fig. 2.— SpeX prism spectra of the esdMs LP 589-7 (esdM5), LSR 0822+1700 (esdM6.5), APMPM 0559-2903 (esdM7) and LEHPM
2-59 (esdM8). Spectra are normalized at their flux peaks and offset by constants (dotted lines). Prominent features in the NIR spectra of
cool esdMs discernible at the resolution of these prism data (λ/∆λ ≈ 150) are labeled. Regions of high telluric opacity are indicated by ⊕
symbols.

LEHPM 2-59 11
Fig. 3.— Optical spectra of the late esdMs LP 589-7 (esdM5), SSSPM 0500-5406 (esdM6.5), APMPM 0559-2903 (esdM7) and LEHPM
2-59 (esdM8) from top to bottom. All data are shifted to their rest frame velocities, normalized at 8100 A˚ and offset by a constant (dotted
lines). Identified molecular features are from Kirkpatrick, Henry, & McCarthy (1991); atomic features are from Kurucz & Bell (1995, see
Table 3). Residual noise from the telluric O2 A-band is indicated by ⊕ symbols.

12 Burgasser & Kirkpatrick
Fig. 4.— Spectral indices CaH2+CaH3 versus TiO5 for dwarfs (points), subdwarfs (open squares) and extreme subdwarfs (filled triangles)
from Hawley, Gizis, & Reid (1996); G97; Gizis & Reid (1997); Reid et al. (2002); LRS03; Le´pine, Shara, & Rich (2004); and Scholz et al.
(2004a). Data from this paper are encircled and labeled, while values for the esdM6.5 LSR 0822+1700 and the sdM9.5 SSSPM 1013-1356
from Le´pine, Shara, & Rich (2004) and Scholz et al. (2004a), respectively, are also labeled. Dashed lines delineate boundaries between
dwarfs, sdMs and esdMs as defined by Eqns. 1 and 2.

LEHPM 2-59 13
Fig. 5.— Comparison of best fit NextGen (top) and AMES Cond (bottom) log g = 5.0 cm s−1 models to the observed red optical
spectrum of LEHPM 2-59 (black). LEHPM 2-59 spectra are normalized at 8100 A˚; model spectra are scaled to their best fit normalization.
The wavelength ranges for which spectral data and models were compared are indicated by the hatched areas.

14 Burgasser & Kirkpatrick
Fig. 6.— Same as Fig. 5 for near infrared spectral fits to LEHPM 2-59. Spectra are normalized at 1.05 µm.

LEHPM 2-59 15
Fig. 7.— Teff s for field M dwarfs (circles) and esdMs (squares) based on spectral model fits. Data are from Kirkpatrick et al. (1993,
SpT ≤ M6); Leggett et al. (1996, 2000, 2001); Gizis (1997, SpT ≤ M6); Schweitzer et al. (1999); Le´pine, Shara, & Rich (2004); and this
paper (oversize symbols). Teff s derived from fits to optical data are indicated by open symbols, those from near infrared data by solid
symbols. The solid line delineates the M dwarf Teff scale from Reid & Hawley (2000); the dashed lines delineate linear fits for esdM Teff s
based on optical (top) and near infrared (bottom) data (Eqns. 10 and 11).

16 Burgasser & Kirkpatrick
TABLE 1
SpeX Observing Log
Source αa δa Jb UT Date tint (s) Airmass Flux Cal. SpT Ref.c
LP 589-7 01h 57m 27.′′92 +01◦ 16m 43.′′3 14.50±0.03 2004 Sep 5 1080 1.13 HD 13936 A0 V 1
LEHPM 2-59 04h 52m 09.′′94 -22◦ 45m 08.′′4 15.52±0.05 2004 Sep 9 720 1.43 HD 32855 A0 V 2
APMPM 0559-2903 05h 58m 58.′′91 -29◦ 03m 26.′′7 14.89±0.04 2005 Dec 31 1440 1.53 HD 41473 A0 V 3
LSR 0822-1700 08h 22m 33.′′69 -17◦ 00m 19.′′9 15.87±0.08 2004 Mar 10 1080 1.01 HD 58383 A0 V 4
References. — (1) Gizis & Reid (1999); (2) Pokorny et al. (2004); (3) Schweitzer et al. (1999); (4) Le´pine, Shara, & Rich (2004)
aJ2000 Coordinates from 2MASS.
bJ magnitudes from 2MASS.
cDiscovery reference for esdM source.
TABLE 2
LDSS-3 Observing Log
Source αa δa Rb UT Date tint (s) Airmass Tell. Cal. SpT Ref.c
LP 589-7 01h 57m 27.′′92 +01◦ 16m 43.′′3 17.39 2005 Dec 4 2100 1.16 HD 603 G2 V 1
LEHPM 2-59 04h 52m 09.′′94 -22◦ 45m 08.′′4 18.72 2005 Dec 4 1800 1.04 HD 31527 G2 V 2
SSSPM 0500-5406 05h 00m 15.′′77 -54◦ 06m 27.′′3 17.42 2005 Dec 4 600 1.50 HD 33967 G2/3 V 3
APMPM 0559-2903 05h 58m 58.′′91 -29◦ 03m 26.′′7 18.08 2005 Dec 4 1200 1.06 HD 33967 G2/3 V 4
References. — (1) Gizis & Reid (1999); (2) Pokorny et al. (2004); (3) Lodieu et al. (2005); (4) Schweitzer et al. (1999).
aJ2000 Coordinates from 2MASS.
bPhotographic R magnitudes from the SuperCosmos Sky Survey.
cDiscovery reference for esdM source.
TABLE 3
Atomic Lines Identified in esdM Optical Spectra
over λλ 6100–8300 A˚
Wavelength Element Elower Eupper Ref.
(A˚) (eV) (eV)
6102.723 Ca I 1.879467 3.910663 1
6122.217 Ca I 1.885935 3.910663 1
6162.173 Ca I 1.899063 3.910663 1
6166.439a Ca I 2.521433 4.531641 1
6169.042a Ca I 2.523157 4.532517 1
6169.563a Ca I 2.525852 4.535043 1
6439.075 Ca I 2.525852 4.450947 1
6449.808 Ca I 2.521433 4.443325 1
6462.567 Ca I 2.523157 4.441254 1
6471.662 Ca I 2.525852 4.441254 1
6493.781a Ca I 2.521433 4.430310 1
6499.650a Ca I 2.523157 4.430310 1
6572.779 Ca I 0.000000 1.885935 1
6717.681 Ca I 2.709192 4.554447 1
7148.150 Ca I 2.709192 4.443325 2
7202.200 Ca I 2.709192 4.430310 2
7326.145 Ca I 2.932710 4.624710 2
7664.911 K I 0.000000 1.617220 1
7698.974 K I 0.000000 1.610064 1
7800.259 Rb I 0.000000 1.589158 3
7947.597 Rb I 0.000000 1.559697 3
8183.255 Na I 2.102439 3.617221 4
8194.790a Na I 2.104571 3.617221 4
8194.824a Na I 2.104571 3.617215 4
References. — (1) Wiese, Smith, & Glennon
(1966); (2) Kurucz (1988); (3) Warner (1968); (4)
Kurucz & Peytremann (1975).
aBlend.

LEHPM 2-59 17
TABLE 4
Alkali Atomic Line Equivalent Widths (A˚).
Source Ca I Ca I K I Rb I Rb I Na I Vrad
(6573 A˚) (7326 A˚) (7665/7699 A˚) (7800 A˚) (7948 A˚) (8183/8195 A˚) (km s−1)
LP 589-7 1.17±0.06 0.76±0.08 9±2 <0.10 0.28±0.03 4.54±0.14 -82±9
SSSPM 0500-5406 2.03±0.15 0.99±0.16 26±2 <0.15 0.5±0.2 6.5±0.4 216±9
APMPM 0559-2903 2.0±0.2 1.1±0.2 27.7±1.9 0.34±0.10 0.62±0.11 6.74±0.14 181±5
LEHPM 2-59 2.17±0.13 0.93±0.12 39.7±1.2 0.60±0.07 0.68±0.05 7.2±0.3 79±9
TABLE 5
Spectral Ratios and Classification.
Source CaH1 CaH2a CaH3a TiO5 Color-Ma SpT
LP 589-7 0.472 0.325 (4.8) 0.493 (4.9) 0.649 1.49 (4.7) sd/esdM5
SSSPM 0500-5406 0.310 0.220 (6.5) 0.331 (7.0) 0.755 1.64 (6.8) esdM6.5
APMPM 0559-2903 0.342 0.217 (6.6) 0.331 (7.0) 0.604 1.67 (7.1) esdM7
LEHPM 2-59 0.267 0.175 (7.3) 0.265 (7.9) 0.656 1.79 (8.7) esdM8
aNumerical esdM subtype in parentheses based on Eqns. 1-3.
TABLE 6
Spectral Model Fits.a
NextGen AMES Cond
5.0b 5.5 5.0 5.5
Source SpT Teff [M/H] Teff [M/H] Teff [M/H] Teff [M/H]
(K) (dex) (K) (dex) (K) (dex) (K) (dex)
Optical
LP 589-7 sd/esdM5 3300 -1.0 3300 -1.0 3200 -1.0 3300 -1.0
SSSPM 0500-5406 esdM6.5 3100 -1.5 3200 -1.5 3000 -1.5 3200 -1.0
APMPM 0559-2903 esdM7 3100 -1.5 3200 -1.0 3000 -1.5 3100 -1.0
LEHPM 2-59 esdM8 3000 -1.5 3100 -1.5 3000 -1.5 3100 -1.0
Near Infrared
LP 589-7 sd/esdM5 3200 -2.0 3300 -1.5 3200 -2.0 3200 -1.5
LSR 0822+1700 esdM6.5 2900 -2.0 3000 -2.0 2900 -2.0 3000 -2.0
APMPM 0559-2903 esdM7 3000 -2.0 3100 -1.5 3000 -1.5 3000 -1.5
LEHPM 2-59 esdM8 2800 -2.0 3000 -2.0 2800 -2.0 3000 -2.0
aValues in bold denote overall best fit models.
bLogarithm of surface gravity in cm s−2.

18 Burgasser & Kirkpatrick
TABLE 7
Properties of LEHPM 2-59.
Parameter Value Ref.
αa 04h 52m 09.′′94 1
δa -22◦ 45m 08.′′4 1
SpT esdM8 2,3,4
RESOb 18.72 5,6
RUSNOb 18.82 7
ISSSb 16.86 5,6
IUSNOb 17.24 7
IDENIS 16.74±0.09 8
J 15.52±0.05 1
H 15.25±0.08 1
Ks 14.76±0.11 1
µ 0.′′746±0.′′016 yr−1 6
θ 174.◦7±1.◦2 6
dest 66±15 pc 2,4
Vrad 79±8 km s−1 2
[U, V,W ]c [135,-180,-80] km s−1 2
Teff 2800-3000 K 2
[M/H] -1.5 to -2.0 2
References. — (1) 2MASS; (2)
This paper; (3) Gizis (1997); (4)
Le´pine, Rich, & Shara (2003a); (5) SSS
(Hambly et al. 2001a,b,c); (6) Pokorny et al.
(2004); (7) USNO-B1.0 (Monet et al. 2003);
(8) DENIS (Epchtein et al. 1997).
a2MASS coordinates, equinox J2000 and
epoch 1998 Nov 29 (UT).
bPhotographic R (IIIaF) and IN (IV-N)
magnitudes.
cAssuming a distance of 66 pc.

This figure "f1.jpg" is available in "jpg"
format from:
http://arXiv.org/ps/astro-ph/0603382v1